Where Are You Talking From?: Estimating the Location of tweets Using Recurrent Neural Networks

ABSTRACT

How can we estimate the location of user-generated content using textual data without location-specific information to understand urban space? Understanding urban space is important to tackle the issues that cities face, e.g. disasters, traffic congestion. Although event information reported with location data on microblog are very informational, many users post them without their locations because of the privacy concerns. To address this issue, some studies have attempted to estimate the location where the users post their tweets by analyzing the text. While those works have introduced various techniques for effective estimation, they have taken a lot of effort to do so. In this paper, we propose an approach that can estimate the location without those efforts. To achieve this goal, we adopt bidirectional Long-Short Term Memory (BLSTM). In our experiment, we use the geotagged tweets that were posted in Japan and treat location estimation as a multi-class classification problem where the location of tweets should be classified into administrative districts. As a result, we show that our proposed method can classify the location of tweets with higher accuracy than baseline methods.

References

1. Cheng, Z., Caverlee, J., and Lee, K. You are where you tweet: a content-based approach to geo-locating twitter users. In Proceedings of the 19th ACM international conference on Information and knowledge management (2010), ACM, pp. 759--768.
2. dos Santos, C. N., and Gatti, M. Deep convolutional neural networks for sentiment analysis of short texts. In COLING (2014), pp. 69--78.
3. Graves, A. Supervised sequence labelling. Springer, 2012.
4. Hochreiter, S., and Schmidhuber, J. Long short-term memory. Neural computation 9, 8 (1997), 1735--1780.
5. Ikawa, Y., Enoki, M., and Tatsubori, M. Location inference using microblog messages. In Proceedings of the 21st international conference companion on World Wide Web (2012), ACM, pp. 687--690.
6. Kingma, D., and Ba, J. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980 (2014).
7. Kinsella, S., Murdock, V., and O'Hare, N. I'm eating a sandwich in glasgow: modeling locations with tweets. In Proceedings of the 3rd international workshop on Search and mining user-generated contents (2011), ACM, pp. 61--68.
8. Kiros, R., Zhu, Y., Salakhutdinov, R. R., Zemel, R., Urtasun, R., Torralba, A., and Fidler, S. Skip-thought vectors. In Advances in Neural Information Processing Systems (2015), pp. 3276--3284.
9. Roller, S., Speriosu, M., Rallapalli, S., Wing, B., and Baldridge, J. Supervised text-based geolocation using language models on an adaptive grid. In Proceedings of the 2012 Joint Conference on Empirical Methods in Natural Language Processing and Computational Natural Language Learning (2012), Association for Computational Linguistics, pp. 1500--1510.
10. Schuster, M., and Paliwal, K. K. Bidirectional recurrent neural networks. Signal Processing, IEEE Transactions on 45, 11 (1997), 2673--2681.
11. Severyn, A., and Moschitti, A. Twitter sentiment analysis with deep convolutional neural networks. In Proceedings of the 38th International ACM SIGIR Conference on Research and Development in Information Retrieval (2015), ACM, pp. 959--962.
12. Tai, K. S., Socher, R., and Manning, C. D. Improved semantic representations from tree-structured long short-term memory networks. arXiv preprint arXiv:1503.00075 (2015).